Friday, January 3, 2025

Flash 1.5, Gemma 2 and Challenge Astra

Share


1.5 Flash excels at summarization, chat functions, picture and video captioning, information extraction from lengthy paperwork and tables, and extra. It is because it’s been educated by 1.5 Professional by a course of referred to as “distillation,” the place probably the most important data and expertise from a bigger mannequin are transferred to a smaller, extra environment friendly mannequin.

Learn extra about 1.5 Flash on the Gemini technology page, and study 1.5 Flash’s availability and pricing. We’ll share extra particulars in an up to date Gemini 1.5 technical report quickly.

Considerably bettering 1.5 Professional

Over the previous couple of months, we’ve considerably improved 1.5 Professional, our greatest mannequin for basic efficiency throughout a variety of duties.

Past extending its context window to 2 million tokens, we’ve enhanced its code era, logical reasoning and planning, multi-turn dialog, and audio and picture understanding by information and algorithmic advances. We see robust enhancements on public and inner benchmarks for every of those duties.

1.5 Professional can now comply with more and more complicated and nuanced directions, together with ones that specify product-level habits involving function, format and elegance. We’ve improved management over the mannequin’s responses for particular use circumstances, like crafting the persona and response type of a chat agent or automating workflows by a number of operate calls. And we’ve enabled customers to steer mannequin habits by setting system instructions.

We added audio understanding within the Gemini API and Google AI Studio, so 1.5 Professional can now motive throughout picture and audio for movies uploaded in Google AI Studio. And we’re now integrating 1.5 Professional into Google merchandise, together with Gemini Advanced and in Workspace apps.

Learn extra about 1.5 Professional on the Gemini technology page. Extra particulars are coming quickly in our up to date Gemini 1.5 technical report.

Gemini Nano understands multimodal inputs

Gemini Nano is increasing past text-only inputs to incorporate pictures as effectively. Beginning with Pixel, functions utilizing Gemini Nano with Multimodality will be capable of perceive the world the way in which individuals do — not simply by textual content, but in addition by sight, sound and spoken language.

Learn extra about Gemini 1.0 Nano on Android.



Source link

Read more

Read More